Estimating student’s performance based on item response theory in a MOOC environment with peer assessment

Peer Assessment is a powerful strategy to support educational activities and the consequent learners’ success. Learning performance of participating is often estimated in a peer assessment setting using Item Response Theory. In this paper, a feasibility of estimating individual performance is examined for a simulated data set representing a MOOC environment, where one thousand students are supposed to perform a Peer Assessment session, where each peer assesses three other peers’ work. For each student the modeling traits “ability”, “consistency”, and “strictness” are evaluated using Generalized Partial Credit Model, and the validity of such calculation is confirmed. While taking into consideration the limits of the synthetic sample production, this experiment provides an evidence of the possibility to predict learning performance in the large scale learning conditions of a MOOC.